MCP server

Server address

https://vouchexplorer.vercel.app/api/mcp

Streamable HTTP, no sign-in, read-only. It reads public data only: public VOUCHED.td files and public GitHub activity.

Connect it

Claude

On claude.ai or the desktop app, open Settings → Connectors, choose Add custom connector, and paste the address above.

Claude Code

Run this in your terminal, outside a Claude Code session:

claude mcp add --transport http --scope user vouch-explorer https://vouchexplorer.vercel.app/api/mcp

• --scope user makes it available in every project.
• --scope project adds it to this project only, in a .mcp.json file you can commit to share it with your team.

Check that it's connected with claude mcp list, or type /mcp in a new Claude Code session to see its status and tools.

Claude asks for permission the first time it calls a tool. To remove it: claude mcp remove vouch-explorer.

ChatGPT

In Settings → Apps & Connectors, turn on developer mode, create a connector, and paste the address. No authentication is needed.

Cursor

Add this to ~/.cursor/mcp.json, or to .cursor/mcp.json in a project:

{
  "mcpServers": {
    "vouch-explorer": { "url": "https://vouchexplorer.vercel.app/api/mcp" }
  }
}

VS Code

Add this to .vscode/mcp.json:

{
  "servers": {
    "vouch-explorer": { "type": "http", "url": "https://vouchexplorer.vercel.app/api/mcp" }
  }
}

Example questions

Once it's connected, ask as you normally would:

• “Has anyone denounced @someone? Who, and with what note?”
• “Who did ghostty-org/ghostty denounce this year?”
• “Where does @pluiedev contribute, and which of those projects vouched for them?”

Tools

trust_lookup

Vouch list entries for a GitHub user or an owner/repo. For a person, every project that vouched for or denounced them. For a repo, everyone its list vouches for or denounces.

• query: a login (mitchellh) or a repo (ghostty-org/ghostty).
• status (optional): vouched or denounced.

Each entry has the project, the status, the maintainer's note, the voucher, the date and a link to the commit that wrote the line. The voucher is the author of that commit. It is null when GitHub doesn't link the commit to an account, and it is never guessed. At most 150 entries come back per call. Use status to narrow a large list.

contributor_activity

A person's public profile, their contributions per repo (commits, pull requests, issues, reviews) with the vouch status where a project lists them, and their recent pull requests, reviews and issues with links.

• login: a GitHub login.

What the chat is told

The server gives every chat the same rules as the AI brief on profile pages:

• Use only the data the tools return. If evidence is missing, say so.
• Don't rate or score the person, and don't recommend vouching or denouncing.
• Link only to URLs the tools return. Never guess an unknown voucher.

Your chat's own model writes the answer. Vouch Explorer doesn't run a model for these calls and doesn't see your conversation, only the tool calls and their arguments.

Limits

Lookups for a person search all of GitHub and can take up to 30 seconds the first time. Results are cached for a few hours after that. Every chat shares one GitHub token, so under heavy use a call can fail on GitHub's rate limit. The tool then says so; try again a minute later.

Try it without a chat

The server speaks JSON-RPC over HTTP, so a plain request works too:

curl https://vouchexplorer.vercel.app/api/mcp \
  -H 'content-type: application/json' \
  -H 'accept: application/json, text/event-stream' \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/call",
       "params":{"name":"trust_lookup","arguments":{"query":"ghostty-org/ghostty","status":"denounced"}}}'

The same data is also at GET /api/trust?q=<login or owner/repo>. The source is on GitHub. See also privacy.